What are key considerations for visualizing data well?

• Keep it simple! (KISS principle)
• Don’t cherry-pick
  • Represent data truthfully
• Make accessible visualizations
  • e.g., color palettes like viridis work for colorblind people

Origin

• ggplot2 did not create “the grammar of graphics.”
• ggplot2 implements the ideas in The Grammar of Graphics (2005) by Leland Wilkinson.

Example data set: food

food <- read_csv("../data/food.csv")

food |>
  select(item, food_group, calories, carbs)

# A tibble: 61 × 4
   item        food_group calories carbs
   <chr>       <chr>         <dbl> <dbl>
 1 Apple       fruit            52 13.8 
 2 Asparagus   vegetable        20  3.88
 3 Avocado     fruit           160  8.53
 4 Banana      fruit            89 22.8 
 5 Chickpea    grains          180 30.0 
 6 String Bean vegetable        31  7.13
 7 Beef        meat            288  0   
 8 Bell Pepper vegetable        26  6.03
 9 Crab        fish             87  0.04
10 Broccoli    vegetable        34  6.64
# ℹ 51 more rows

Example scatter plot

Observations represented by dots:

food <- read_csv("../data/food.csv")

food |>
  ggplot() +
  geom_point(aes(x = calories, y = carbs))

Example text plot

Observations represented by the item label:

food |>
  ggplot() +
  geom_text(aes(x = calories, y = carbs, label = item))

Example bar plot (more complex)

Observations represented by bars:

food |>
  # filter for high cholesterol foods:
  filter(cholesterol > 50) |>
  ggplot() +
  # set nice x axis label:
  xlab("Food") +
  # set nice y axis label:
  ylab("Cholesterol (mg)") +
  # set bar chart
  geom_col(aes(
    # sort bars by descending amount of cholesterol
    x = reorder(item, -cholesterol),
    y = cholesterol,
    # set bar color (fill) by food_group
    fill = food_group
  )) +
  # tilt labels for readability:
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

Syntax review

With ggplot, you can combine multiple layers to create simple or complex data visualizations. In general terms, the structure is:

data |>
  ggplot() +
  geom_...(aes(x = ..., y = ...)) +
  ...

Why hard-code aesthetics?

• Sometimes you want to set aesthetics that are not tied to the data.
• For example, you might want to set the color of all points to be green.
• Or you might want to set the size of all points to be huge.

Example scatter plot with fixed aesthetics

food |>
  ggplot() +
  # note that the color green goes outside of aes():
  geom_point(aes(x = calories, y = carbs, size = 10), color = "green")